Illumination-robust face recognition using ridge regressive bilinear models

The performance of face recognition is greatly affected by illumination changes because intra-person variation of the captured images under different lighting conditions can be much bigger than the inter-person variation. This paper proposes an illumination-robust face recognition by separating an identity factor and an illumination factor using symmetric bilinear models. The translation procedure in the bilinear model requires a repetitive computation of matrix inverse operations to reach the identity and illumination factors. This computation may result in a non-convergent case when the observation has noisy information or the model is overfitted. To alleviate this situation, we suggest a ridge regressive bilinear model that combines the ridge regression into the bilinear model. This provides a number of advantages: it stabilizes the bilinear model by shrinking the range of identity and illumination factors appropriately and improves the recognition performance. Experimental results show that the ridge regressive bilinear model significantly outperforms other existing methods such as the eigenface, quotient image, and the bilinear model in terms of the recognition rate under a variety of illuminations.